Zapier and Make · Course lab · about 120 minutes · 6 tasks · marked out of 100, pass at 60

Run a clinic's appointment requests without anyone copying anything

The situation

Dr Saira's physiotherapy clinic in Lahore takes appointment requests through a website form and WhatsApp. The receptionist copies each one into a sheet, decides whether it is a new patient, a follow-up or a complaint, and forwards it. Requests that arrive after hours wait until the next afternoon.

Build the automation that captures, sorts and routes every request, uses AI only where it helps, and tells someone when it breaks.

What you'll be able to show

  • Build a trigger-and-actions automation from a real test submission
  • Use filters, paths and formatting so the output is usable downstream
  • Add an AI classification step with a safe fallback path
  • Turn on error alerts and prove them with a deliberate failure

What you need

  • Zapier (free plan) or Make (free plan); either is fine
  • A Google Form or Tally form, Google Sheets and Gmail
  • The AI step built into the tool, or an OpenAI or Gemini connection

Tasks

  1. 1Capture one real request

    Build the form with Name, Phone, Message and Preferred day. Create the trigger and test it with one real submission so you can see the exact fields.
    A correct result: A screenshot of the trigger test showing every field as it arrives.
  2. 2Write it down cleanly

    Add a Sheets step that appends Name, Phone, Message, Preferred day and Received at. Add a formatting step so the phone is stored as 03XX-XXXXXXX and names in title case.
    A correct result: Three test rows with clean phone numbers and names, no matter how they were typed.
  3. 3Sort with AI, safely

    Add an AI step that reads the message and returns exactly one word: NEW, FOLLOWUP, COMPLAINT or UNSURE. Write the instruction so any other output counts as UNSURE.
    A correct result: Six test messages and the label each received, including one that is deliberately ambiguous.
  4. 4Route by label

    Add paths (Zapier) or a router (Make): NEW to reception, FOLLOWUP to the treating physio, COMPLAINT to Dr Saira, and UNSURE to reception with the word CHECK in the subject.
    A correct result: Four messages reaching four different inboxes, and the run history showing each path.
  5. 5Filter early

    Add a filter as early as possible that stops spam (no phone number, or a message under five characters) before any AI or email step runs.
    A correct result: A spam test that stops at the filter, with no AI call and no email.
  6. 6Know when it breaks

    Turn on error notifications. Then break it on purpose (rename the sheet tab) and submit once. Fix it and confirm the next run succeeds.
    A correct result: The error notification, and the next successful run in the history.

What to hand in

Screenshots of the trigger test, the clean sheet, the six labels, the four routed messages, the stopped spam test and the error alert; plus a one-paragraph note for the receptionist on what she no longer has to do.

How it is marked

CriterionPoints
Built from a real submission and writes clean, formatted rows 20
AI step returns one label with an UNSURE fallback 25
Paths route each label correctly 20
Spam filtered before any AI or email step 15
Error alert proven with a deliberate failure 20
Total · pass at 60 100

Hand in your lab

Create a free BvLogic ID to hand in your lab, get it marked, and have it on your certificate.

Create your BvLogic ID